Demo Outdoor Store

A shopping assistant that sells through product cards and comparisons, remembers sizes, and is open to shopping agents.

The problem it solves

Stores want conversational shopping that converts: showing products instead of describing them, remembering what a customer wears, confirming totals before charging, and being findable by the AI shopping agents customers increasingly send on their behalf.

Products, comparisons and the cart appear as interactive components in the conversation
Orders are placed only after the shopper confirms the total, and never twice
External shopping agents can search the catalog over MCP, A2A and /llms.txt

Run it on your computer

git clone https://github.com/karawash-projects/amoxi.gitcd amoxinpm installnpm run buildnpm run dev -w shopping-assistant

Then open http://localhost:8083. The tests run with npm test -w shopping-assistant.

Read the source and README on GitHub

Amoxi features in this example

Feature	Why it matters here
Generative UI, React included	The product carousel is a React component the agent shows with validated props; the rest of the store ships little JavaScript.
Memory	Sizes and preferences are remembered across visits and recalled when relevant.
Tools with approval	Placing an order and starting a return wait for the shopper's confirmation.
Knowledge base (RAG)	Gear guides and the returns policy answer "which bag for -10°C?" style questions.
Agentic commerce	Search and comparison tools are offered to outside agents over MCP and A2A, and every product page has markdown and JSON versions.
Cached pages with revalidation	Catalog pages are served from cache and refreshed when an order changes stock.